lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Otis Gospodnetic <otis_gospodne...@yahoo.com>
Subject Re: federated / meta search
Date Sat, 19 Jun 2010 02:26:10 GMT
Lance, which project in Solr are you referring to?


Thanks,

Otis
----Sematext :: http://sematext.com/ :: Solr - Lucene - Nutch
Lucene ecosystem search :: http://search-lucene.com/



----- Original Message ----
> From: Lance Norskog <goksron@gmail.com>
> To: solr-user@lucene.apache.org
> Sent: Fri, June 18, 2010 8:16:46 PM
> Subject: Re: federated / meta search
> 
> Yes, you can do this. You need to have a common system for creating
unique 
> ids for the documents.

Also, there's an odd problem around relevance. 
> Relevance scoring is
based on all of the terms in a field in the whole index, 
> and there is
a "statistical fingerprint" of this for an index. With two 
> indexes
from two sources, the terms in the documents will not have the 
> same
"fingerprint". Relevance scores from one shard will not match 
> the
meaning of a document's score in the other shard.

There is a 
> project to make this work in Solr, but it is not nearly finished.

Lance 
> Norskog

On Fri, Jun 18, 2010 at 4:28 AM, Sascha Szott <
> ymailto="mailto:szott@zib.de" href="mailto:szott@zib.de">szott@zib.de> 
> wrote:
> Hi Joe & Markus,
>
> sounds good! Maybe I should 
> better add a note on the Wiki page on federated
> search 
> [1].
>
> Thanks,
> Sascha
>
> [1] 
> href="http://wiki.apache.org/solr/FederatedSearch" target=_blank 
> >http://wiki.apache.org/solr/FederatedSearch
>
> Joe Calderon 
> wrote:
>>
>> yes, you can use distributed search across shards 
> with different
>> schemas as long as the query only references 
> overlapping fields, i
>> usually test adding new fields or tokenizers 
> on one shard and deploy
>> only after i verified its working 
> properly
>>
>> On Thu, Jun 17, 2010 at 1:10 PM, Markus 
> Jelsma<
> href="mailto:markus.jelsma@buyways.nl">markus.jelsma@buyways.nl>
>> 
>  wrote:
>>>
>>> 
> Hi,
>>>
>>>
>>>
>>> Check out 
> Solr sharding [1] capabilities. I never tested it with
>>> different 
> schema's but if each node is queried with fields that it 
> supports,
>>> it should return useful 
> results.
>>>
>>>
>>>
>>> [1]: 
> href="http://wiki.apache.org/solr/DistributedSearch" target=_blank 
> >http://wiki.apache.org/solr/DistributedSearch
>>>
>>>
>>>
>>> 
> Cheers.
>>>
>>> -----Original 
> message-----
>>> From: Sascha Szott<
> ymailto="mailto:szott@zib.de" 
> href="mailto:szott@zib.de">szott@zib.de>
>>> Sent: Thu 
> 17-06-2010 19:44
>>> To: 
> ymailto="mailto:solr-user@lucene.apache.org" 
> href="mailto:solr-user@lucene.apache.org">solr-user@lucene.apache.org;
>>> 
> Subject: federated / meta search
>>>
>>> Hi 
> folks,
>>>
>>> if I'm seeing it right Solr currently 
> does not provide any support for
>>> federated / meta searching. 
> Therefore, I'd like to know if anyone has
>>> already put efforts 
> into this direction? Moreover, is federated / meta
>>> search 
> considered a scenario Solr should be able to deal with at all or
>>> 
> is it (far) beyond the scope of Solr?
>>>
>>> To be more 
> precise, I'll give you a short explanation of my
>>> requirements. 
> Assume, there are a couple of Solr instances running at
>>> 
> different places. The documents stored within those instances are 
> all
>>> from the same domain (bibliographic records), but it can not 
> be ensured
>>> that the schema definitions conform to 100%. But lets 
> say, there are at
>>> least some index fields that are present in 
> all instances (fields with
>>> the same name and type definition). 
> Now, I'd like to perform a search on
>>> all instances at the same 
> time (with the restriction that the query
>>> contains only those 
> fields that overlap among the different schemas) and
>>> combine the 
> results in a reasonable way by utilizing the score
>>> information 
> associated with each hit. Please note, that due to legal
>>> issues 
> it is not feasible to build a single index that integrates the
>>> 
> documents of all Solr instances under 
> consideration.
>>>
>>> Thanks in 
> advance,
>>> 
> Sascha
>>>
>>>
>
>



-- 
> 
Lance Norskog

> href="mailto:goksron@gmail.com">goksron@gmail.com

Mime
View raw message